Secure systems are systems that limit access in one or more ways. Secure systems may include physical secure systems, including but not limited to automated teller machines (ATMs), electronically locked doors, and/or other physical secure systems, as well as virtual secure systems, including but not limited to online access to a particular website, an account on a website, a particular area within a website, and/or other virtual secure systems. One commonly employed security measure for access to secure systems is a personal identification number, or PIN. Known examples include a bankcard or debit card that requires a particular personal identification number before a (financial) transaction can be initiated and/or completed.
Some secure systems employ speech recognition technology. Speech recognition commonly uses one or more recorded audio files from a particular speaker that include sound generated by that speaker. Through analysis and/or signal processing, a speech recognition system may determine which word or phrase the particular speaker uttered, and/or whether the utterance matches a predetermined passphrase, password, and/or other utterance that has been predetermined to grant a speaker access to the secure system.
Some secure systems employ voice recognition technology (also referred to as speaker verification technology). Speaker verification commonly uses recorded audio files that include sound generated by one or more people (also referred to herein as speakers). Based on one or more audio files, one or more audio characteristics and/or parameters may be assessed per speaker. A set of audio characteristics, together with any other information that may be useful to characterize and/or identify a particular speaker, may form the basis of a model of the particular speaker (for example, a parameterized model may represent a particular speaker). Models of speakers may be used to assess, estimate, determine, and/or otherwise reach a decision on whether a particular recorded audio file (or a particular speaker model) appears to represent a known speaker, and/or which known speaker in particular.
In some implementations, speakers may be represented by speaker models including sets and/or vectors or values and/or coefficients. By way of non-limiting example, some speaker verification systems may be based on speaker models that include fixed-length vectors. As used herein, a fixed-length vector may refer to a vector having the same number (or a similar number) of values and/or coefficients regardless of the number and/or length of the recorded audio files associated with the particular speaker represented by the particular model that includes the fixed-length vector.